Hillslope stability problems experienced in Three Gorges Reservoir Area after the water impoundment in the reservoir have been widely studied in the literature. However, the contributions of morphometric, meteorological and seismic variables altering hillslope deformations and the variation in their roles over time is yet to be explored. This study aims at addressing this gap in the literature. To do so, we generate hillslope deformation time series using Interferometric synthetic aperture radar (InSAR) techniques and calculate mean line-of-sight velocities for eight time windows between March 2017 and April 2021. For each of these time windows, we build a random forest model and regress six variables (slope steepness, distance to river, total precipitation, snow cover, earthquake intensity and territorial water storage) against mean line-of-sight velocities. For each of these models, we quantify the variable importance. Our findings show that earthquakes and precipitation have the highest contribution to the surface deformations, across different time windows. Additionally, we run a suite of bivariate analyses to assess the contribution of reservoir water level, an informative layer whose variability is only measured across time. We show that hillslopes mainly close to the reservoir exhibit strong negative correlation between surface deformation and reservoir water level. Overall, our finding could be considered as a step towards developing a predictive tool to identify expected hillslope deformation in the future.

中文翻译：

---

三峡库区山坡变形控制变量的探讨


三峡库区蓄水后所经历的山坡稳定性问题已被文献广泛研究。然而，形态、气象和地震变量对山坡变形的贡献及其随时间的变化仍有待探索。本研究旨在解决文献中的这一空白。为此，我们使用干涉合成孔径雷达 (InSAR) 技术生成山坡变形时间序列，并计算 2017 年 3 月至 2021 年 4 月之间八个时间窗口的平均视线速度。对于每个时间窗口，我们构建一个随机森林模型并根据平均视线速度回归六个变量（坡度、到河流的距离、总降水量、积雪、地震强度和领土蓄水量）。对于每个模型，我们量化变量的重要性。我们的研究结果表明，在不同的时间窗口内，地震和降水对地表变形的贡献最大。此外，我们还进行了一套双变量分析来评估水库水位的贡献，水库水位是一个信息层，其变化仅随时间变化而测量。我们发现，主要靠近水库的山坡地表变形与水库水位之间表现出很强的负相关性。总的来说，我们的发现可以被视为开发预测工具以识别未来预期的山坡变形的一步。